printf white space character in bash script

Question

How can the white space character be escaped in printf for the purpose of this script that currently runs in a Cygwin client? The line containing the whitespace (which by itself works from the command line as echo $SPC) is

SPC=$(printf \\$(printf '%03o' 32))

In case you might wonder why, in this code I am trying to simulate an associative array such that it is not dependent on Bash 4+...would like this to be a generic example that theoretically can be run in virtually any bash environment.

Here is the entire script:

#!/bin/bash

DEFAULT_INDEX=""
LAST_INDEX=0
SPC=$(printf \\$(printf '%03o' 32))

# MICROSOFT OUTLOOK VERSIONS.
OUTLOOKS=( "Cancel^Do$SPCnot$SPCtry$SPCto$SPCsend$SPCmessage$SPCvia$SPCOutlook"
           "URL^Just$SPCthe$SPCmailto$SPCURL"
           "Outlook$SPC2003^C:\Program$SPCFiles\Microsoft$SPCOffice\Office12"
           "Outlook$SPC2007^C:\Program$SPCFiles\Microsoft$SPCOffice\Office12"
           "Outlook$SPC2010^C:\Program$SPCFiles\Microsoft$SPCOffice\Office14" )

# ITERATES THROUGH ARRAY AND MAKES NUMBERED SELECTION LIST.
printOutlooksArraySelectionList() {
    LIST_TITLE="MICROSOFT OUTLOOK SELECTION LIST"
    echo
    printf "\t\t\E[37;1;44m%-0s %s %s %s\033[0m" $LIST_TITLE
    echo
    echo
    DEFAULT_VALUE="Outlook 2010"
    index=0
    for i in "${OUTLOOKS[@]}"; do
        key=${i%%^*}
        value=${i##*^}
        if [ "$key" == "$DEFAULT_KEY" ]; then
            DEFAULT_INDEX=$index
    fi
            printf "\t\E[37;44m%-3s %-13s\033[0m \E[1;34;40m%-41s\033[0m\n" \
                   $index $key $value
            echo
            ((index++))
    done
    # SETS LAST_INDEX TO INDEX VALUE OF LAST ITEM IN ARRAY.
    ((LAST_INDEX = index - 1))
}

printOutlooksArraySelectionList

Would like the output to look like the following:

            MICROSOFT OUTLOOK SELECTION LIST

    0   Cancel        Do not try to send message via Outlook

    1   URL           Just the mailto URL

    2   Outlook 2003  C:\Program Files\Microsoft Office\Office12

    3   Outlook 2007  C:\Program Files\Microsoft Office\Office12

    4   Outlook 2010  C:\Program Files\Microsoft Office\Office14

To work-around, I ended up not using printf, but instead:

echo -e "\t\E[37;44m${index}\t${key}\033[0m\t\t\E[1;34;40m${value}\033[0m"

adequate for my purposes with this script.

Answer 1

I don't understand what you're trying to do. Your code does set SPC to a string of length 1 whose single character is a space. This is a highly convoluted way of doing it: SPC=" " would work just as well and be comprehensible.

I haven't reviewed your whole script, but I do see something wrong later down: you write things like "URL^Just$SPCthe$SPCmailto$SPCURL". This produces the string URL^Just, because the variables $SPCthe, $SPCmailto and $SPCURL are not defined. When a variable expansion is immediately followed by a character that could be part of the variable name, you need to indicate the end of the variable name, for example by putting the name into braces: URL^Just${SPC}the${SPC}mailto${SPC}URL. I don't understand why you're not writing "URL^Just the mailto URL", though.

For your encoding of small hash tables, I suggest using two array, once for keys and one for values. That way, you can store arbitrary strings, without fearing encoding issues. Here's a sketch of what lookup and addition can look like (taking the names of the keys and values arrays as parameters is left as an exercise):

lookup () { # $1=key; set value to the value found
  for ((i=1; i<=${#keys}; i++)); do
    if [[ ${keys[$i]} = $1 ]]; then value=${values[$i]}; return 0; fi
  done
  unset value; return 1
}
add () { # $1=key $2=value
  for ((i=1; i<=${#keys}; i++)); do
    if [[ ${keys[$i]} = $1 ]]; then values[$i]=$2; return; fi
  done
  keys[$i]=$1; values[$i]=$2
}

A more efficient method would be to store each entry in a shell variable. For example, if the table foo maps somekey to somevalue, then set the variable table_foo_somekey=somevalue. You need to encode the keys into alphanumerics, which is non-trivial (one method is to hash the keys, e.g. with sha1, and store the key in table_foo_${sha1_of_somekey}_key and the value in table_foo_${sha1_of_somekey}_key). If you need to enumerate the keys in a table, you need to store the list separately.

Or, when you need things that go beyond the capabilities of a simple shell, you could reach for a more powerful tool. Ksh93 (not pdksh), bash 4 and zsh all have associative arrays. Beyond that, there's Perl and Python and the rest of them.

Answer 2

• Since you've already used quotes around the strings, using $SPC in them is not necessary and gives you no benefit. You don't need any definition of space there.

• As Gilles pointed out, strings like "URL^Just$SPCthe$SPCmailto$SPCURL" contain non-existing variables. A variable ends either at a character that cannot be used in any variable name, or when it is (properly) delimiterred with {} as in ${key}.

Answer 3

I've already mentioned some misunderstanding of that problem. But anyway.

The line containing the whitespace (which by itself works from the command line as echo $SPC) is

SPC=$(printf \\$(printf '%03o' 32))

Actually, that line do not containing any whitespaces. It expands in

printf \\040

which expands in nothing good. And printfing a printf output is a really bad idea at all.

Print could show space symbols if you want it:

$> printf '  %03o\n' 32
  040

Leading nulls and leading zeros too:

$> printf "%010o\n" 32
0000000040

$> printf "% 10o\n" 32
        40

UPD by comment:

$> printf '\\%03o\n' 32
\040